Local AI Registry

Recipe

minimax-m2-7-ud-iq4-xs-dgx-spark-gb10-128gb-llama-cpp-tp1

minimax-m2-7-ud-iq4-xs-dgx-spark-gb10-128gb-llama-cpp-tp1
JSON APIFind compatible recipes

Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
b8833 (45cac7ca7)
Accelerators
1
Tensor parallel
1
Context tokens
32,768
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/MiniMax-M2.7
Repository
unsloth/MiniMax-M2.7
Status
known
Link type
Exact Hub repository

Public Hugging Face repository confirmed by the Hub API.

Candidate: useful compatibility or speed evidence. The registry does not offer Run until promotion requirements are met.

Observed configuration

llama.cpp

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cmoi6yfxh000nk304si0uyyp5

Observed source tokens

Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.

  1. /opt/llama.cpp/build-cuda/bin/llama-server
  2. --jinja
  3. -fa
  4. on
  5. --no-warmup
  6. -t
  7. 20
  8. -c
  9. 32768
  10. -ctk
  11. q8_0
  12. -ctv
  13. q8_0
  14. -np
  15. 1
  16. -a
  17. MiniMax-M2.7-UD-IQ4_XS
  18. -m
  19. /models/MiniMax-M2.7-GGUF/UD-IQ4_XS/MiniMax-M2.7-UD-IQ4_XS-00001-of-00004.gguf
  20. --host
  21. 127.0.0.1
  22. --port
  23. 8012
  24. --spec-type
  25. ngram-simple
  26. --draft-max
  27. 16
  28. --draft-min
  29. 1
  30. --draft-p-min
  31. 0.5
  32. --spec-ngram-size-n
  33. 4
FlagValue
-faon
-t20
-c32768
-ctkq8_0
-ctvq8_0
-np1
-aMiniMax-M2.7-UD-IQ4_XS
-m/models/MiniMax-M2.7-GGUF/UD-IQ4_XS/MiniMax-M2.7-UD-IQ4_XS-00001-of-00004.gguf
--host127.0.0.1
--port8012
--spec-typengram-simple
--draft-max16
--draft-min1
--draft-p-min0.5
--spec-ngram-size-n4

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,768—31783observedminimax-m2-7-ud-iq4-xs-dgx-spark-gb10-128gb-llama-cpp-tp1-sweep

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

hardware count
1
hardware id
dgx-spark-gb10-128gb
id
minimax-m2-7-ud-iq4-xs-dgx-spark-gb10-128gb-llama-cpp-tp1
model instance id
unsloth-minimax-m2-7--ud-iq4-xs
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
minimax-m2-7-ud-iq4-xs-dgx-spark-gb10-128gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
b8833 (45cac7ca7)

serving

max concurrency
1
max context tokens
32,768
tensor parallel
1
Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at
2026-08-30T09:26:09Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
capabilities.reasoning · provenance · captured at
2026-08-30T09:26:09Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
capabilities.tools · provenance · captured at
2026-08-30T09:26:09Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
capabilities.vision · provenance · captured at
2026-08-30T09:26:09Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
engine.graph mode · provenance · captured at
2026-08-30T09:26:09Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
serving.kv cache tokens · provenance · captured at
2026-08-30T09:26:09Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown

metadata

localmaxxing · backend
cuda
localmaxxing · hardware label
GB10 Grace Blackwell
localmaxxing · notes
MiniMax-M2.7-UD-IQ4_XS (unsloth dynamic Q4, 101 GB GGUF) on a single DGX Spark with llama.cpp ngram-simple speculative decode: draft-max=16, draft-min=1, p-min=0.5, ngram-size=4. Sherlock prose, pp=128, tg=128, c=1, n=29 warm. tokSOut = 30.98 t/s MEDIAN (std 4.08). 1.255× over AR baseline (24.68 t/s). Tested all 5 llama.cpp ngram variants — ngram-simple won, ngram-mod 2nd, ngram-cache regressed below AR, the two ngram-map-* variants crashed. peakVramGb=107 via free -g (101 GB GGUF + KV + overhead in 128 GiB UMA, basically no headroom). Note: clean reproductions on different Sparks produce medians anywhere from 30–42 t/s (std 4–17) — fitting a 101 GB model in 128 GiB UMA leaves the kernel paging weights opportunistically and you catch some 'fast' windows depending on luck. Submitting the most-reproducible cell. Repo: github.com/my-other-github-account/spark-bench-reproducers/tree/main/minimax-m27-ngram-spec-spark. By @banana_baeee.
localmaxxing · run id
cmoi6yfxh000nk304si0uyyp5
localmaxxing · tokenized · fidelity
faithful

provenance

captured at
2026-08-30T09:26:09Z

sources

captured atkindurl
2026-08-30T09:26:09Znormalized-recipewww.localmaxxing.com/en/runs/cmoi6yfxh000nk304si0uyyp5 ↗